The traditional model of payments, where a person initiates a purchase and a bank or card network processes it, is on the cusp of a significant transformation. Emerging trends in the banking sector indicate that software agents powered by artificial intelligence (AI) may soon take on the crucial role of initiating financial transactions on behalf of users.

Leading this charge is Visa, which is actively rolling out its “Agentic Ready” programme across Europe. This initiative is designed to test how existing financial systems can effectively handle transactions initiated by AI agents. Visa is collaborating with major banks, including Commerzbank and DZ Bank, to prepare the current payment infrastructure for a future where AI software can autonomously search for products, make purchasing decisions, and complete transactions according to user-defined goals or rules. This means an AI agent, rather than a human customer, could confirm and execute a purchase, marking a fundamental shift from the current process.

A core challenge in this evolution lies in how transactions begin. Current payment systems are meticulously built around human identity and intent, requiring verification that a person has authorized a purchase. If AI agents become the initiators, banks will need to develop sophisticated new methods to confirm identity and intent at a system level. This includes establishing how an AI agent proves it is acting legitimately on behalf of a user and defining the extent of autonomy it should possess. Visa’s model envisions software agents handling routine or repetitive purchases with minimal human intervention, guided by user-set rules. For instance, an AI system could monitor supply levels and compare prices, then automatically complete a transaction once specific conditions are met.

The scale of this anticipated change is comparable, according to Visa, to the early adoption of online payments, which necessitated significant adaptations within banking systems. Banks involved in these early trials, such as Commerzbank and DZ Bank, are rigorously testing the practical integration of AI agents into their existing frameworks without compromising stringent compliance rules. This includes crucial checks related to fraud detection, maintaining comprehensive audit trails, and securing explicit customer consent. Given the highly regulated nature of these areas, any alteration to transaction initiation must strictly adhere to current oversight standards. A RepRisk report has already highlighted that banks are facing increasingly frequent and costly issues linked to AI, potentially leading to multi-million-dollar losses.

Visa's primary focus in this endeavor is on the underlying infrastructure, rather than developing consumer-facing tools. The company is dedicated to defining how payment networks should operate when the 'customer' is a piece of software. This encompasses establishing protocols for agent authentication, transaction approval mechanisms, and clear processes for handling disputes should issues arise. Beyond individual transactions, AI agents hold significant potential for enterprise purchasing, particularly in large organizations where procurement often involves multiple, time-consuming approval steps. AI agents could streamline and compress this process by managing routine purchases within predefined limits, thereby reducing manual workload. However, this also necessitates that companies implement clear and robust rules governing the actions of these agents to mitigate risks of errors or misuse.

More broadly, financial institutions are heavily investing in AI to automate back-office operations and achieve cost efficiencies, with many reorganizing teams to prioritize data and AI strategy. Concurrently, regulators are intensifying their scrutiny of AI’s role in decision-making, particularly in sensitive areas like credit assessment and fraud detection. These collective developments suggest that payments could emerge as one of the first sectors where AI agents operate with substantial autonomy. While banks will retain responsibility for setting rules, monitoring activities, and managing exceptions, the day-to-day act of initiating a transaction may, in certain contexts, require less direct human input. Visa’s current efforts are centered on rigorous testing and system design, as the financial infrastructure must evolve to accommodate a new type of 'user'—one that may not hold a physical card but possesses the capability to complete a purchase.